Spatial Estimation of Soil Total Nitrogen Using Cokriging with Predicted Soil Organic Matter Content

Abstract

Accurate measurement of soil total N (TN) content in agricultural fields is important to guide reasonable application of nitrogenous fertilizer. Estimation of soil TN content with limited in situ data at an acceptable level of accuracy is important because laboratory measurement of N is a time‐ and labor‐consuming procedure. This study was conducted to evaluate cokriging of soil TN with predicted soil organic matter (SOM) content as auxiliary data. The SOM content was predicted by cokriging with a digital number (DN) of Band 1 of Landsat Enhanced Thematic Mapper (ETM) imagery. Soil TN content was estimated by using 88 soil samples for prediction and 43 soil samples for validation in a study area of 367 km2 in Haining City, China. Field‐measured soil TN content ranged from 0.47 to 2.48 g kg−1, with a mean of 1.25 g kg−1 Soil TN content of all 131 soil samples including samples for prediction and validation was highly correlated with measured (r = 0.81, p < 0.01) and predicted (r = 0.81, p < 0.01) SOM content in paddy fields. Then, the predicted SOM content was used as auxiliary variable for the prediction of soil TN content. By using the 43 samples for validation, we had a mean error (ME) of 0.03 g kg−1 and a root mean square error (RMSE) of 0.31 g kg−1 for kriging, and a mean error of 0.00 g kg−1 and a root mean square error of 0.25 g kg−1 for cokriging, respectively. Our results indicate cokriging with predicted SOM content data was superior to kriging. In addition, predicted data of the auxiliary variable have the potential to be useful for cokriging when the predicted auxiliary data have high prediction accuracy.

Spatial interpolators for improving the mapping of carbon stock of the arboreal vegetation in Brazilian biomes of Atlantic forest and Savanna

“…The estimation of non-sampled points through co-kriging was described by Yamamoto and Landim (2013), among others, which is an extension of kriging, where one or more easier measured variables help to explain the target variable as they are correlated (Wu et al, 2009). For this approach, the identification of the neighbors' weight was made with the fitted parameters of the crossed semivariogram.…”

Prediction of the Soil Organic Matter (SOM) Content from Moist Soil Using Synchronous Two-Dimensional Correlation Spectroscopy (2D-COS) Analysis

“…Note that the four experimental configurations (A–D) are designed to illustrate the effect of the SOM prediction accuracies when different combinations of spectral bands have been used for the analysis. Specifically in experiment (D), where only the SOM featured bands have been selected like that of the traditional method commonly adopted by workers in the field [ 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 , 60 , 61 , 62 ], the experiment is designed to compare the impacts when SOM only bands are utilised with respect to the proposed band selection method as set out in Experiment (C).…”

“…Apart from the correction approach as outlined in the previous paragraph, alternative methodology has been the discovery of the SOM related spectral bands [ 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 ] particularly those which are not greatly affected by the absorption feature of moisture in the soil [ 60 , 61 , 62 ]. There are numerous reports about the SOM-related absorption features in the vis-NIR spectral range and studies have revealed that spectral bands in the 800–1400 nm, 1600–1700 nm, 2100–2200 nm and 2300–2500 nm region [ 61 ] are relatively free of interference because of the presence of moisture content in the soil.…”

Hypsometric approaches to Eucalyptus experiments in Brazil

“…In this approach, weight was obtained with the parameters of the semivariogram fitted for the characteristic being evaluated (HT). The estimation of unsampled points made from co-kriging was described by Yamamoto and Landim (2013), among other authors, and is an extension of kriging, in which one or more easy-tomeasure variables helps explain a difficult-to-measure primary variable (Wu et al, 2009). For this approach, the identification of weights was obtained with the parameters of crossed semivariance fitted for the variable HT, where the secondary variable was the DBH, which was obtained for all the individuals in the stand.…”